Financial Datasets

What it is

A REST API for US company data — financial statements, SEC filings, insider and institutional ownership, earnings, operating KPIs, index-fund holdings, news and equity prices — sold to people wiring a language model into the market. The OpenAPI spec lists 56 paths. The agent framing is not only marketing: an agent can POST to a signup endpoint to open its own account, then mint a Stripe link and hand it to the human who pays.

The SEC-derived datasets are parsed by the vendor straight from EDGAR with no intermediary, which is the reason to prefer this over a reseller — every figure traces to a filing, and both standardised and as-reported line items are served. Equity prices are the exception: those are licensed from Databento, which the docs name outright.

Pricing

Three self-serve plans and nothing free. $20 buys 1,000 one-time credits that never expire; Build is $200 a month for 100,000 requests; Scale is $2,000 a month for 1,000,000 and is the only plan carrying redistribution rights, webhooks and an uptime SLA. Annual billing is ten months for twelve. Overage runs $10 per 1,000 requests on Build and $5 on Scale, capped at 1 million and 10 million requests a month.

Four dataset families — IPOs, operating KPIs, forward guidance and non-GAAP metrics — are metered as multiples rather than gated: each request counts as 8 on Credits, 4 on Build, 2 on Scale.

Data & coverage

27,000-plus US tickers, active and delisted, with fundamentals back to the early 1990s. History is tiered, and the headline number is the generous half: core history is 1 year on Credits and 30+ years on Build and Scale, while price history is 1 year on Credits and 3+ years on both paid plans. "30+ years" means the filings, not the charts.

Prices are end-of-day — the historical endpoint offers day, week, month and year intervals and nothing finer. A real-time last-price snapshot exists, per ticker and for the whole market, but there are no intraday bars, no ticks and no streaming socket. Webhooks push filing, earnings and insider events within seconds, on Scale and Enterprise only.

Integrations

The hosted MCP server is the flagship surface. The docs advertise "30+ tools" and name 27, which cover the REST endpoints including the screener, and list setup for Claude, Claude Code, Cursor, Codex, ChatGPT, Hermes and OpenClaw.

There is also an MIT-licensed reference server in the vendor's GitHub organisation, financial-datasets/mcp-server — 2,288 stars, and not the product. Its last commit on main is 5 June 2025, it has no tagged release, and it exposes a fraction of the hosted server's tools. Nothing from it reaches a package registry, and the names a reader would guess belong to other people: pip install financial-datasets fetches the founder's unrelated 2024 library for generating synthetic question-and-answer sets from 10-Ks, and financial-datasets-mcp-server on PyPI is a stranger's empty 0.1.0 placeholder.

There is no first-party client library in any language — plain HTTP with an X-API-KEY header, plus llms.txt files for anyone generating their own.

Limitations

• US equities only. Non-US markets, options, indices and currencies are listed as unavailable on the vendor's own coverage page, and no plan changes that.
• No free tier and no trial. $20 is the floor, and an unfunded key returns 402 on every call.
• End-of-day prices, so this is not a feed. No intraday bars, no ticks, no WebSocket.
• Price history stops at 3+ years even on Scale, against 30+ years of fundamentals.
• Redistribution is a $2,000-a-month feature. Below Scale you may build commercially on the data but not pass it to your own users.
• Crypto endpoints still answer, but were removed from the docs, the OpenAPI spec and the product pages during 2026. Treat them as gone.
• Allowances move sharply — Build fell from two million requests to 100,000 inside three months, and overage rose fiftyfold — so a budget built on last quarter's numbers will be wrong.

Alternatives

For SEC filings and XBRL fundamentals over MCP with no bill at all, SEC EDGAR MCP wraps EDGAR itself and costs only the compute, at the price of doing the normalisation yourself. Financial Modeling Prep and Intrinio sell overlapping US fundamentals with cheaper entry points and no first-party MCP server. If what you want is intraday or tick prices, this is the wrong shelf — see the market data APIs. The rest of the category is at MCP servers and AI agents.